Many organisations are developing their own machine learning algorithms to predict the results of this year’s World Cup. There’s nothing quite like the World Cup to bring about superstition in fans, ranging from hanging lucky flags to following the word of a psychic octopus. But can we use big data and machine learning to accurately foresee the winners of this year’s tournament, and who is the machine’s choice?
Goldman Sachs have developed their own algorithm, based upon data mining team statistics and breaking down the performance of individual players. They ran the model 200,000 times, with an additional million variations included to predict the results of every game. Goldman predicts that Brazil will be returning from Moscow as the victors, beating Germany in the finals.
Researchers from the Technische Universitat of Dortmund, Technical University of Munich and Ghent University have come together to produce their own algorithm which they ran 100,000 times. Unlike Goldman Sachs, they also took into account the population size of the country, GDP, bookies’ odds and how many of the national team players play regularly together at league level.
This model predicts Spain to win, ranked higher than Germany by only two percent. The researchers suggest this is due to Germany having a much tougher road on its way to the finals.
South African data firm Principa predicts that Germany are the most likely winners. They developed three different models using a variety of techniques all of which reached similar conclusions.
So far in the tournament the vast majority of Principa’s individual scores have been wrong, but they have been fairly accurate in deciding winners and losers. Before writing them off, it is important to note that Principa have quite the track record, successfully predicting the outcomes of the 2015 Rugby World Cup and the 2016 Oscars.
Finally we reach the models developed by the start-up Unanimous AI, who’s record of predictions are tough to beat: ranging from the Oscars, the Kentucky Derby to even Time Person of the Year.
Their algorithms are normally used to garner public opinion and market research, however when turning its eye to the World Cup, Germany is predicted as the winner. The company uses swarm technology to accumulate public opinion to predict outcomes, as the ultimate ‘ask the audience’.
Before you start placing any money down, do take into account that while machine learning can be incredibly accurate, a sport like football still has so many unpredictable variables ranging from injuries to referees.
Here at K2 we have a huge network of machine learning professionals, eager to take on the challenges of a constantly evolving world. Or are you a machine learning expert who is looking for the next big project? If so why not get in touch?
Dominic Whaley, Digital Content Specialist based at K2 Partnering Solutions HQ in London.